PjParam should have minimal pub methods

[nothingmuch]

only the receiver_pubkey method needs to be pub (because payjoin-cli uses the value), but ohttp_keys, expiration and endpoint only need to be accessed internally to the payjoin crate after the URI has been built

[DanGould]

One part of this I'm still not completely happy with is the visibility of the getter methods on v2::PjParams. receiver_pubkey is used in payjoin-cli as a session-id parameter. The v2::PjParams::endpoint method is only used in payjoin-ffi, and I can't help but wonder if that's a mistake vestigial of the old way we were indexing sessions by endpoint before receiver_pubkey was used for indexing, as in payjoin-cli, related to #336. The other public getters for ohttp_keys and expiration are unused and can be made more private, though I'd like to give them holistic consideration the context of all of the getters.

Originally posted by @DanGould in #1048 (comment)

[DanGould]

In #995 we decided to use receiver_pubkey as the index. Upon reflection @nothingmuch mentioned that the comparison with Urls might be a better user experience even if receiver_pubkey is used to disambiguate. Yuval may you shed some light on how you'd like this fully realized?

[nothingmuch]

comparing by RK should be done first, and payjoin-cli should error if that's ever duplicated.

as a special case, we can check if the full URI is identical if that happens (ideally by comparing the URI as a string, not just the fields we recognize in our extras struct) and soften the error to instruct the sender to use the resume command instead of just complaining about a duplicate RK

[nothingmuch]

closed by #1082